We open source 100 common financial questions and a training dataset for use within the PEER framework, containing 100 sets of inputs and outputs of the Planning Agent and the Expressing Agent. The dataset is applicable to agent training, pre-training, and supervised fine-tuning in various financial analysis scenarios.
For more training data, please provide your email via GitHub Issues, we typically respond within 2 days.
Due to compliance issues, the relevant data of the Executing agent is still in the approval process, please wait patiently.
We have collected 100 professional questions from our business scenarios and categorized them into nine distinct categories, including information query, general financial QA, report interpretation, target analysis, strategy advice, major events interpretation, macro analysis, market analysis, policy interpretation.
Each set of data contains the input and output of the Planning Agent. The input is the original query proposed by the user. The output is the sub-questions generated by the agent, serving as the interpretation framework.
Each set of data contains the input and output of the Expressing Agent. The input is the answer to each qub-question given by the Executing Agent through collecting information. The output is the professional answer after synthesis and rewriting.